arXiv subjects

Robin Dupont

Publications and source records attributed to Robin Dupont.

4 recordsLinked to original sources

arXiv2026

Exploring the Relaxation Landscape of a 2D Quantum Magnet on a 256-Qubit Processor

How quantum matter relaxes far from equilibrium is a central open problem in many-body physics, and one for which analog quantum simulators are well positioned to move from confirming theory to discovering new physics. Here, we use a two-dimensional Rydberg atom array of 256 qubits to map the relaxation landscape of the two-dimensional transverse-field Ising model across its phase diagram. Beyond the expected rapid thermalization, we identify two further regimes. The first is a prethermal regime whose dynamics are governed by an effective XY model. The second, and most unexpected, is a crossover regime characterized by a slowdown in relaxation. This slowdown occurs precisely where state-of-the-art classical tensor-network methods lose control at late times, whereas the quantum simulation remains consistent across system sizes. These results establish Rydberg atom arrays as a platform for scientific discovery in nonequilibrium quantum many-body dynamics.

quant-ph↗

arXiv2026

One-to-one quantum simulation of a frustrated magnet with 256 qubits

Analog quantum simulators offer a powerful microscopic probe of quantum many-body systems, yet have largely been benchmarked against model Hamiltonians rather than real materials. Here, we use a 256-qubit Rydberg simulator to implement the effective Hamiltonian of the frustrated triangular-lattice magnet TmMgGaO$_4$. Simulated magnetization curves agree quantitatively with susceptibility measurements on single crystals, and both platforms consistently determine the antiferromagnetic phase transition. Snapshot-resolved analysis confirms that quantum fluctuations, rather than disorder, govern the intermediate paramagnetic regime. Having established this correspondence, we access non-equilibrium dynamics following a sudden quench, a regime at picosecond material timescales where entanglement growth places the problem beyond classical reach. The simulator reveals thermalization of local observables, demonstrating that analog quantum simulation can reproduce and extend the physics of a real material.

quant-ph↗

arXiv2022

Extracting Effective Subnetworks with Gumbel-Softmax

Large and performant neural networks are often overparameterized and can be drastically reduced in size and complexity thanks to pruning. Pruning is a group of methods, which seeks to remove redundant or unnecessary weights or groups of weights in a network. These techniques allow the creation of lightweight networks, which are particularly critical in embedded or mobile applications. In this paper, we devise an alternative pruning method that allows extracting effective subnetworks from larger untrained ones. Our method is stochastic and extracts subnetworks by exploring different topologies which are sampled using Gumbel Softmax. The latter is also used to train probability distributions which measure the relevance of weights in the sampled topologies. The resulting subnetworks are further enhanced using a highly efficient rescaling mechanism that reduces training time and improves performance. Extensive experiments conducted on CIFAR show the outperformance of our subnetwork extraction method against the related work.

cs.CV↗

arXiv2021

Weight Reparametrization for Budget-Aware Network Pruning

Pruning seeks to design lightweight architectures by removing redundant weights in overparameterized networks. Most of the existing techniques first remove structured sub-networks (filters, channels,...) and then fine-tune the resulting networks to maintain a high accuracy. However, removing a whole structure is a strong topological prior and recovering the accuracy, with fine-tuning, is highly cumbersome. In this paper, we introduce an "end-to-end" lightweight network design that achieves training and pruning simultaneously without fine-tuning. The design principle of our method relies on reparametrization that learns not only the weights but also the topological structure of the lightweight sub-network. This reparametrization acts as a prior (or regularizer) that defines pruning masks implicitly from the weights of the underlying network, without increasing the number of training parameters. Sparsity is induced with a budget loss that provides an accurate pruning. Extensive experiments conducted on the CIFAR10 and the TinyImageNet datasets, using standard architectures (namely Conv4, VGG19 and ResNet18), show compelling results without fine-tuning.

cs.CV↗